Contribution of morphometry in the differential diagnosis of fine-needle thyroid aspirates.

BACKGROUND

Cytologic discrimination of cellular nodules, follicular adenoma, and follicular carcinoma in the thyroid is problematic. Methods are needed to achieve a reliable diagnosis. Some sophisticated tools, such as microarrays, offer great potential but lack accompanying morphologic information.

METHODS

One hundred twelve samples obtained from patients with lesions histopathologically diagnosed as nodular goiter, follicular adenoma, follicular carcinoma, and papillary carcinoma were used. Eight geometric features, such as nuclear area and circular form factor, were measured. The dataset was divided into six overlapping groups to represent the frequently encountered situations in routine practice. Multivariate analysis of variance, Tukey's honestly significant differences test, and discriminant analysis were performed. Statistical analysis was carried out with two conceptually different approaches. In the first, data from all measured nuclei were used. In the second, a subset of data representing the most extreme values of variables was extracted from the entire dataset to simulate the "selection procedure" performed during conventional morphologic examination.

RESULTS

When the selected dataset instead of data from all measured nuclei was used, the correct classification rates in discriminant analysis improved considerably.

CONCLUSIONS

Morphologic examination is based primarily on selection. Using data obtained from all of the cells in morphometry may cause a dilution effect in diagnostically important features. Morphometric studies may also be planned with a proper selection "bias." This may be particularly helpful when isolated abnormal cells carry most of the diagnostic information.